AN INTELLIGENT ONLINE SCHEME BASED K-NEAREST NEIGHBOR CLASSIFIER FOR GEAR SYSTEM FAULT DIAGNOSIS AND CLASSIFICATION

Auteurs :  I. Attoui, B. Oudjani, M.S. BOUAKAZ, K. CHETTAH, W. RICHI
Année : 2018
Domaine : Génie électrique
Type : Communication
Conférence: International Conference on Applied Analysis and Mathematical Modeling ICAAMM 2018
Résumé en PDF :  (résumé en pdf)
Fulltext en PDF : 
Mots clés : Data Analysis, Gear System, Fault Diagnosis, classification, Linear Sequence Discriminant Analysis.

Résumé : 

Gear system faults in all sorts of rotating machines can result in partial or total breakdown,destruction and even catastrophes. By implementation of an adequate fault detection system, thereliability, productivity, safety and availability of the gearboxes that are used to transfer rotating powersource to other devices, provide speed, and torque conversions can be improved. In this paper, aspecific interest is carried to correctly identify type, location, and class of different defects that canappear in the gear and bearing in a gear system with various combinations under different speeds andloads. The solution is based on a three-step algorithm. The first step, based on the Wavelet PacketTransform WPT and FFT algorithms, is used to extract the features of the different sub-bandsfrequencies in the vibration signal from a gear system. Then, in the second step, a dimensionalityreduction based Linear Sequence Discriminant Analysis LSDA algorithm is conducted to reducecomputational overhead for diagnosis and to improve classification performance. Finally, the reducedfeatures were used as the input to a k-nearest neighbor classifier to evaluate the system diagnosisperformance. Based PHM Data Challenge, the experimental results obtained from real gear systemvibration signals for eight different health gearbox conditions demonstrated that the proposed methodis effective in both feature extraction, feature reduction, and also fault classification.